Hugging Face's logo

avatarcontinues
/
blip1000captions

Image-Text-to-Text
Transformers
Safetensors
blip
Inference Endpoints
Model card Files Community
avatarcontinues commited on
Commit
a4e027a
•
1 Parent(s): 257538d

Upload BlipForConditionalGeneration

Browse files
Files changed (2) hide show
  1. config.json +7 -2
  2. model.safetensors +2 -2
config.json CHANGED
@@ -1,5 +1,5 @@
1
  {
2
- "_name_or_path": "Salesforce/blip-image-captioning-base",
3
  "architectures": [
4
  "BlipForConditionalGeneration"
5
  ],
@@ -10,6 +10,7 @@
10
  "model_type": "blip",
11
  "projection_dim": 512,
12
  "text_config": {
 
13
  "initializer_factor": 1.0,
14
  "model_type": "blip_text_model",
15
  "num_attention_heads": 12
@@ -18,9 +19,13 @@
18
  "transformers_version": "4.36.0.dev0",
19
  "vision_config": {
20
  "dropout": 0.0,
 
21
  "initializer_factor": 1.0,
22
  "initializer_range": 0.02,
 
23
  "model_type": "blip_vision_model",
24
- "num_channels": 3
 
 
25
  }
26
  }
 
1
  {
2
+ "_name_or_path": "Salesforce/blip-image-captioning-large",
3
  "architectures": [
4
  "BlipForConditionalGeneration"
5
  ],
 
10
  "model_type": "blip",
11
  "projection_dim": 512,
12
  "text_config": {
13
+ "encoder_hidden_size": 1024,
14
  "initializer_factor": 1.0,
15
  "model_type": "blip_text_model",
16
  "num_attention_heads": 12
 
19
  "transformers_version": "4.36.0.dev0",
20
  "vision_config": {
21
  "dropout": 0.0,
22
+ "hidden_size": 1024,
23
  "initializer_factor": 1.0,
24
  "initializer_range": 0.02,
25
+ "intermediate_size": 4096,
26
  "model_type": "blip_vision_model",
27
+ "num_attention_heads": 16,
28
+ "num_channels": 3,
29
+ "num_hidden_layers": 24
30
  }
31
  }
model.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:1889b2243db6b0e51ef8bf7688f382177ae1920c1379166e3032e9bb11b65116
3
- size 989717056
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3c74c1af33186ac00fd6790f5230614c6929cbaeff5b1fa24691eaefc6ddb21a
3
+ size 1879010480